Proficiency of Nucleic Acid Tests for Avian Influenza Viruses, Australasia

An avian influenza quality assurance program was used to provide information for laboratories on the sensitivity and specificity of their avian influenza nucleic acid testing. Most laboratories were able to correctly detect clinically relevant amounts of influenza virus (H5N1), and results improved as each subsequent panel was tested.


H ighly pathogenic avian infl uenza (HPAI) virus (H5N1)
is endemic among the world's wild bird populations and continued to spread during 2006 to poultry across Asia, Africa, and mainland Europe (1,2). Sensitive, specifi c diagnostic methods are essential for early accurate detection of HPAI virus in the prepandemic and early pandemic phases in countries where no cases have been recorded, such as Australia (3).

The Study
We report results from an avian infl uenza quality assurance program (QAP) that used an established, Internetbased quality assurance reporting system (www.rcpaqap. com.au/serology), allowing remote data entry, rapid result dissemination, and expert comment. The QAP provided feedback to laboratories on NAT characteristics (PCR accuracy, sensitivity, and specifi city), reporting optimization, and assessment of continuously updated laboratory-developed NAT methods.
During 2006, three panels of specimens were distributed to 29 participating laboratories: 15 from Australia (including 4 veterinary laboratories); 2 from Hong Kong Special Administrative Region, People's Republic of China; 5 from Singapore; 1 from New Caledonia; 1 from Malaysia; and 5 from New Zealand. The panels consisted of an Indonesian and a Vietnamese strain of avian infl uenza virus (H5N1), originally isolated from humans and grown in MDCK cells. Viral copy numbers were estimated by comparing real-time RT-PCR crossing-point values to a standard curve generated by using plasmid standards; the amplicon was cloned into pGEMT-Easy (Promega, Madison, WI, USA). Plasmid standard concentrations were estimated as described previously (13) and as recommended by the LightCycler manufacturer (Roche, Indianapolis, IN, USA). Sensitivity of NATs was determined with a range of clinically relevant nucleic acid concentrations of both infl uenza (H5N1) strains (10 3 to 10 -1 copies/μL) to enable laboratories to assess limit of detection (LOD) of their assays. Specifi city was assessed by inclusion of other infl uenza strains and a negative control ( Table 1). All strains and MDCK cells were inactivated by exposure to 50 KGy of γ-irradiation, except for strain A (H7N4), which was inactivated by the addition of lysis buffer (14).
Four experiments to defi ne optimum conditions were conducted. 1) LOD determinations, with a dilution series of all strains, were tested by using real-time RT-PCR (15). 2) Transport media were compared by using serial dilutions of inactivated infl uenza virus (H5N1) in phosphate-buffered saline with gelatin (with antimicrobial agents) (PBSG), TE buffer, and buffer RLT (lysis) (QIAGEN, Valencia, CA, USA), placed at -80°C, -20°C, +4°C, +25°C, and +37°C for 10 days. Each day, 1 tube at each temperature condition was removed, and viral DNA was extracted by using the QIAamp viral RNA minikit (QIAGEN) and tested with a real-time RT-PCR (15). 3) For further stability testing, a test panel diluted in PBSG was sent by courier from Sydney to Hong Kong, held by Australian customs for 7 days, and returned unopened 15 days after dispatch. The temperature range the shipped panel was exposed to is unknown; however, previous temperature loggers have recorded temperatures from 22°C to 33°C. The panel that traveled was tested against a panel that had been stored optimally (-80°C) (15). No difference was detected in the amount of virus in the specimens that traveled compared with optimally stored specimens, indicating that the specimens were stable under normal transport conditions (results not shown). 4) Homogeneity was established before distribution by having the For each panel the samples were diluted in PBSG and transported by courier at ambient temperature in 3 seasons (autumn, winter, summer). Participants were not required to use a certain NAT method. Participants were asked for information on methods used, including extraction and RT-PCR protocol and primer/probe sequences. A total of 780 results were analyzed, and a report was issued to participants within 3 weeks of each survey closing, well before the next panel shipment. This allowed participants to adjust their testing procedures if necessary before the next survey began. Results were reported by participants as positive, negative, or equivocal. For simplicity, we report equivocal results as positive, given that participating laboratories retest an equivocal result and generally do not report such results as negative. On average for the 3 panels, 2.6% of results were reported as equivocal.
Panel 1 contained 8 specimens, which included 2 dilutions (10 3 and 10 1 copies/μL) of each subtype H5N1 strain (Table 1). Only 35% of participants correctly identifi ed all samples; 95% reported a correct result for the highest concentration of both subtype H5N1 strains (10 3 copies/μL); 70% could detect the lower concentration (10 1 copies/μL). Only 46% of participating laboratories used an infl uenza A matrix assay as well as a specifi c H5 assay ( Table 2). Laboratories were advised to use both methods in tandem to reduce the chance of missing variant infl uenza (H5N1) strains that might not be detected by their specifi c H5 assay. Some false positives (1.3%) were reported, and some confusion regarding terminology occurred: many laboratories reported results as for subtype H5N1 assays, despite most of these results being specifi c for the H5 gene only.
For panel 2, no participants correctly identifi ed all samples because of the addition of 2 extremely dilute samples of infl uenza virus (H5N1) (10 0 and 10 -1 copies/μL) that were below the LOD for most laboratories. Eleven percent of participants detected 1 strain of HPAI virus (H5N1) by using primers specifi c for H5 or subtype H5N1 at 1 of the 2 highest dilutions, but not both. In our experience, dilute specimens are useful for assessing the LOD of the testing system because they may highlight the most sensitive methods available. The number of laboratories using a generic infl uenza A test, in addition to a specifi c H5 test, increased to 73% (Table 2).
For panel 3, sensitivity of detection improved compared with panel 2: 25% of participants detected a strain of infl uenza virus (H5N1) at the lowest concentrations. Sensitivity of H5/H5N1 testing for the infl uenza (H5N1) Vietnamese strain increased over time, while sensitivity of *These dilutions were included in panels 2 and 3 only; NA, not applicable. †For panel 3, the dilution for the influenza (H5N1) Vietnamese strain was changed to 1:1,000,000, with a copy number of 5 × 10 0 L. testing decreased slightly over the 3 panels for the infl uenza (H5N1) Indonesian strain (Table 2). Laboratories had altered primer/probe sets to increase sensitivity for the Vietnamese strain, which resulted in decreased sensitivity for the Indonesian strain. Sensitivities of other testing methods (infl uenza A, B, H3) increased during subsequent testing of each panel (data not shown); the number of correct results reported by participants using infl uenza A matrix testing rose from 84% in panel 1 to 91% in panel 3.

Conclusions
Most participants did not disclose their primer/probe sequence information, which made it diffi cult to recommend the most sensitive methods to other participants. However, during a prepandemic phase, having a range of primers/probes being used may be optimal, providing infl uenza A matrix detection is also conducted and QA is maintained, until WHO recommends a method to detect new pandemic strains.
Participants in the avian infl uenza QAP made clear improvements in the sensitivity and specifi city of their NAT methods over time. It is important to provide continuing QA to expose inconsistencies in results or primers that may be skewed toward a particular strain.